$Z$-sight: Unifying Vision-Language-Action and Latent World Modeling
We introduce $Z$-sight, a unified robot foundation model that couples vision-language-action (VLA) and generative latent world modeling. Video-based world action models (WAMs) show the value of joint future-and-action generation for control, yet inherit the heavy cost of video generators. Recent latent-space policies avoid pixel generation, yet typically align intermediate features or rely on frozen target encoders, decoupling future prediction from the VLM's native feature space. $Z$-sight instead predicts the future directly in the visual token space in which its VLM perceives the present. Coupling a foresight expert and a continuous action expert with the VLM in a mixture-of-transformers, $Z$-sight jointly denoises future spatial tokens and actions at inference without pixel reconstruction or external encoders. Trained on human video and robot trajectories, $Z$-sight achieves 81.9% success on SIMPLE (+9.6 points over the strongest baseline) and 73.3% across six Unitree G1 tasks (+20.0 points) at 110 ms per chunk, while remaining competitive on tabletop benchmarks.
Authors
- Nassim A. Samad
- Benjamin J. Kolligs (ORCID: https://orcid.org/0000-0002-6393-237X)
- Sasanka Nagavalli
- Jyotirmoy V. Deshmukh (ORCID: https://orcid.org/0000-0003-4683-5540)
- Gaurav S. Sukhatme (ORCID: https://orcid.org/0000-0003-2408-474X)
- Ruohai Ge (ORCID: https://orcid.org/0009-0007-3978-3320)
- Muchen Xu
- Celina Shiyu Wang
- Rong Xue
- Aditya Dutt
- Junjie Ye
- Hao Dong
- Yiqi Zhao
- Yue Wang
- Yuliang Guo
Publication Details
- Journal
- arXiv (Cornell University)
- Published
- 2026-11-18
- DOI
- https://doi.org/10.48550/arxiv.2611.24940
- Primary Topic
- Robot Manipulation and Learning
- Type
- preprint